Resampled Datasets Are Not Enough: Mitigating Societal Bias Beyond Single Attributes
Abstract
We tackle societal bias in image-text datasets by removing spurious correlations between protected groups and image attributes. Traditional methods only target labeled attributes, ignoring biases from unlabeled ones. Using text-guided inpainting models, our approach ensures protected group independence from all attributes and mitigates inpainting biases through data filtering. Evaluations on multi-label image classification and image captioning tasks show our method effectively reduces bias without compromising performance across various models.
Authors
- Yusuke Hirota
- Jerone Andrews
- Dora Zhao*
- Orestis Papakyriakopoulos*
- Apostolos Modas
- Yuta Nakashima*
- Alice Xiang
*External Authors
Venue
EMNLP 2024
Date
2024